vllm v0.30.0
vllm v0.30.0 is a new release of the open-source vllm inference/serving engine providing Python wheels and Docker images across CUDA, ROCm, XPU and CPU targets. The release adds broad model support, GPU weight-caching for faster engine restarts, watermarking support, host-resident sparse decode tiers, and many performance and speculative-decoding improvements.
What it does
Developers can use it to serve supported models through an inference API on their own hardware.
- Availability
- Released September 22, 2026; Python wheels and Docker images listed
- Pricing model
- Not stated.
- Deployment
- Python wheels and Docker images for CUDA, ROCm, CPU, and XPU
- Privacy
- Not stated.
- Licence
- Not stated.
Capability
Modality
Access
Deployment
Pricing
Target user
Evidence
6 claims- vLLM 0.30.0 was released on September 22. [1]
- The release provides Python wheels and Docker images for CUDA, ROCm, CPU, and XPU. [1]
- The release adds support for DeepSeek-V4.1-Flash, GLM-5.3-Flash, K2-Horizon, and other models. [1]
GLM-5.3-Flash : EPLB support ( #55119 ), FlashKDA for KDA chunked prefill (1.7-3.8x faster than the Triton chunk path, #55737 ), dense and masked-MHA sparse prefill for the NoPE (256, 0, 256) layout ( #55738 ), decode hot-path cleanups removing per-layer copies and a duplicate router GEMM ( #55736 ), and the glm5next NVIDIA subtree now packaged in wheels ( #55214 ).
GitHub watchlist. Quoted text is outside CC BY 4.0 and keeps its original rights. - The release adds Gumbel-max watermarked generation and detection, with per-request opt-out. [1]
Watermarking : Gumbel-max watermarked generation and detection with a keyed PRF, per-request opt-out and an example detection endpoint ( #54053 ); dual-key Gumbel-max makes it compatible with speculative decoding ( #56122 ); the Rust frontend forwards the per-request controls ( #56338 ).
GitHub watchlist. Quoted text is outside CC BY 4.0 and keeps its original rights. - HiSparse can spill sparse-MLA KV pages to pinned host memory under GPU pressure. [1]
HiSparse : a host-resident tier for sparse-MLA decode that spills KV pages to pinned host memory under GPU pressure and serves top-k misses from a per-request GPU hot buffer, enabled through HiSparseConnector ( #53781 ), with Prometheus counters ( #56061 ), a host cache shared across TP ranks ( #56629 ), and the attention config inferred from the connector ( #57041 ).
GitHub watchlist. Quoted text is outside CC BY 4.0 and keeps its original rights. - Model Runner V2 adds dual-batch overlap in eager mode and with FULL CUDA graphs. [1]
Sources
1
1
GitHub watchlistOwner-controlled
https://github.com/vllm-project/vllm/releases/tag/v0.30.0
Retrieved 2026-10-28 00:02 UTC
Verified
Versions
35| Version | Effective | Change | Changed fields | State |
|---|---|---|---|---|
| 35 | 2026-10-28 00:02 UTC | material change | capability, modality, access, availability, deployment | Current |
| 34 | 2026-10-26 01:12 UTC | material change | capability, access, availability, deployment | Current |
| 33 | 2026-10-25 01:09 UTC | material change | capability, deployment, availability, deployment | Current |
| 32 | 2026-10-24 01:40 UTC | material change | modality, access, availability, deployment | Current |
| 31 | 2026-10-23 01:34 UTC | material change | capability, modality, access, deployment, pricing, availability, deployment | Current |
| 30 | 2026-10-22 00:52 UTC | material change | capability, deployment, pricing, availability, deployment, privacy | Current |
| 29 | 2026-10-20 19:44 UTC | material change | capability, access, deployment, availability, pricing model, deployment | Current |
| 28 | 2026-10-19 20:40 UTC | material change | capability, modality, deployment, pricing, availability, deployment | Current |
| 27 | 2026-10-18 21:38 UTC | material change | capability, modality, deployment, availability, deployment | Current |
| 26 | 2026-10-17 22:29 UTC | material change | capability, pricing, availability | Current |
| 25 | 2026-10-16 23:28 UTC | material change | capability, modality, access, deployment, pricing, availability, deployment | Current |
| 24 | 2026-10-16 00:25 UTC | material change | capability, modality, access, availability, deployment | Current |
| 23 | 2026-10-15 01:17 UTC | material change | capability, modality, access, deployment, availability, deployment | Current |
| 22 | 2026-10-14 02:09 UTC | material change | capability, access, deployment, availability, deployment | Current |
| 21 | 2026-10-13 03:01 UTC | material change | capability, availability, deployment | Current |
| 20 | 2026-10-12 03:58 UTC | material change | capability, modality, access, pricing, availability, deployment | Current |
| 19 | 2026-10-11 04:51 UTC | material change | capability, modality, access, pricing, availability | Current |
| 18 | 2026-10-10 05:45 UTC | material change | capability, access, availability | Current |
| 17 | 2026-10-09 06:40 UTC | material change | capability, modality, access, availability, deployment | Current |
| 16 | 2026-10-08 07:35 UTC | material change | capability, availability, deployment | Current |
| 15 | 2026-10-07 08:32 UTC | material change | capability, modality, access, deployment, availability, deployment | Current |
| 14 | 2026-10-06 09:31 UTC | material change | capability, modality, access, deployment, availability, deployment | Current |
| 13 | 2026-10-05 10:28 UTC | material change | capability, access, availability, deployment | Current |
| 12 | 2026-10-04 11:24 UTC | material change | capability, availability, deployment | Current |
| 11 | 2026-10-03 12:22 UTC | material change | capability, modality, availability, deployment | Current |
| 10 | 2026-10-02 13:17 UTC | material change | capability, access, availability, deployment | Current |
| 9 | 2026-10-01 14:09 UTC | material change | capability, modality, access, availability, deployment | Current |
| 8 | 2026-09-30 15:03 UTC | material change | capability, access, availability, deployment | Current |
| 7 | 2026-09-29 15:57 UTC | material change | capability, modality, access, pricing, availability, deployment | Current |
| 6 | 2026-09-28 16:47 UTC | material change | capability, pricing, availability, deployment | Current |
| 5 | 2026-09-27 17:42 UTC | material change | capability, modality, access, pricing, availability, deployment | Current |
| 4 | 2026-09-26 18:36 UTC | material change | capability, modality, access, pricing, availability, deployment | Current |
| 3 | 2026-09-25 19:29 UTC | material change | capability, access, pricing, availability, deployment | Current |
| 2 | 2026-09-24 20:22 UTC | material change | capability, deployment, pricing, availability, deployment | Current |
| 1 | 2026-09-23 21:18 UTC | publication | None | Current |
In briefs
1- New model releases span quantized language models and image generation Daily brief 2026-09-23
Original catalog content is licensed under CC BY 4.0, attributed to wowsuchai. Source quotations and linked content keep their original rights.
